Some Properties of a Measure of Information Discrepancy  

作　　者：FANGShun-lan FANGWei-wu 

机构地区：[1]DepartmentofFinance,RenminUniversityofChina,Beijing100872,China [2]InstituteofAppliedMathematics,ChineseAcademyofSciences,Beijing100080,China

出　　处：《Systems Science and Systems Engineering》2002年第4期397-408,共12页系统科学与系统工程学报（英文版）

基　　金：Thiswork issupported by National Natural Science Foundation of China(90 10 30 33,39830 0 70 ) andNKBRSF (G19980 30 60 )

摘　　要：Based on a group of axioms, a measure of information discrepancyamong multiple information sources has been introduced in and it possesses some peculiar properties compared with other measures of information discrepancy, so it can be used in some areas, where the traditional measures are not valid or not efficient, for example, in the study of DNA sequence comparison, prediction of protein structure class, evidence analysis, questionnaire analysis, and so on. In this paper, using the optimization techniques, we prove that it is a distance function and show that it is also an approximation of χ2 function. These two properties will stimulate further applications of the measure to information processing and system analysis.Based on a group of axioms, a measure of information discrepancyamong multiple information sources has been introduced in and it possesses some peculiar properties compared with other measures of information discrepancy, so it can be used in some areas, where the traditional measures are not valid or not efficient, for example, in the study of DNA sequence comparison, prediction of protein structure class, evidence analysis, questionnaire analysis, and so on. In this paper, using the optimization techniques, we prove that it is a distance function and show that it is also an approximation of χ2 function. These two properties will stimulate further applications of the measure to information processing and system analysis.

关 键 词：information discrepancy DISTANCE χ2 function Shannon entropy K-L entropy 

分 类 号：G202[文化科学—传播学] O236[理学—运筹学与控制论]